Most AI chatbots avoid telling users who to vote for, but a new study reveals significant political biases when forced to choose between candidates. When asked to recommend a candidate for the upcoming midterms, SpaceXAI's Grok 4.7 selected the Republican candidate in 92% of responses, whereas OpenAI's GPT-6.1 Sol picked the Democrat in 97%. Across nearly 500 races, the two models agreed on their choices less than 10% of the time.
Despite a low adoption rate for election information—only 8% of Americans use chatbots for this, according to Pew Research—a study by Convovo suggests the models powering these answers are not neutral. The researchers, led by Cornell University professor Tapan Parikh, found that when asked directly via API, the models refused to answer in consumer apps but provided recommendations in developer interfaces.
"ChatGPT is designed to be objective by default, including in political conversations, and shouldn't have political bias in any direction," stated an OpenAI spokesperson. The spokesperson clarified that the API tests, which involved prompts explicitly demanding a candidate recommendation, do not reflect the user experience of most people, who are typically directed to decline such requests. SpaceXAI did not respond to the study in time for comment.
Parikh noted that while consumer apps initially refused, he was able to jailbreak them with simple prompts to reveal their decision-making processes. He emphasized that these biases might surface in broader conversations, stating, "If people are using these LLMs not just to directly ask who to vote for, but are having these more general conversations, then these biases are going to come up."
The study highlighted the differing sources the models cited. Grok frequently surfaced conservative sources alongside nonpartisan sites, including a Christian conservative voter guide in 39% of answers and Turning Point Action in 23%. Conversely, GPT leaned heavily on Wikipedia, the Associated Press, Axios, and Reddit, where OpenAI holds content deals. Parikh suspects both models are searching for justification to support pre-existing leans, though Grok made an exception when the Republican candidate was involved in controversy.
In a surprising twist, when internet search was disabled, the models relied on different mechanisms. GPT still demonstrated a Democratic lean based on its training weights, picking the Democrat in 24 out of 30 races. However, Grok's Republican lean disappeared; it refused to pick in 27 out of 30 races, and when it did choose, it picked the Democrat. Parikh concluded that for GPT, the decision is pre-existing in its weights, while for Grok, the search terms and sources determine the outcome.
This study arrives just weeks before the November 3 midterms and adds fuel to the debate over political bias in AI. As President Donald Trump has attacked "woke AI," and SpaceXAI CEO Elon Musk has donated over $445 million to GOP candidates, the political stakes are high. OpenAI CEO Sam Altman has a history of donating to Democrats, but recent shifts suggest a complex landscape of AI funding and bias.
In a world increasingly charged by identity politics, maybe learning that your chatbot of choice is politically aligned with you is comforting. But for those who still want the opportunity to review each candidate's stance and make their own choices, here are some resources:
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.